dotfiles

設計や計画の壁打ちインタビュー。候補案、推奨案、選択式UIを出さず、事実と開かれた問いだけを1つずつ投げて、ユーザー自身に設計理由を言語化させる。行き詰まりにはヒントを段階的に強め、明示的なギブアップ時のみ一般的に多数派の回答を出す。 ユーザーが「壁打ち」「kabeuchi」と言ったとき、または設計判断を自分の言葉で説明できるようになりたい・定着させたいという意図を示したときに使う。推奨案付きで決めたい場合は grill-me の領分なので、この skill は使わない。

Learning support coach for communicating in English on GitHub Issues and Pull Requests. Provides structure and stock phrases, then gives hints on the user's own draft instead of writing English for them. Assumes CEFR A2. Use when the user says「英語で返信したい」「英語でPRを書きたい」「この英語の意味を教えて」"english coach", or asks for help writing or reading English on Issues / PRs / OSS.

Control herdr from inside it. Manage workspaces and tabs, split panes, spawn agents, read output, and wait for state changes — all via CLI commands that talk to the running herdr instance over a local unix socket. Use when running inside herdr (HERDR_ENV=1).

日本語の技術文書における論証、演出の抑制、整形の規範。段落単位の再配置・削除・橋渡しで論証の筋を通す編集も扱う。技術ブログ記事、design doc、ポストモーテムを書くとき、推敲するとき、または議論の筋を点検して直すときに使用する。

Append a 日次掘り下げ section to today's AI diary. Run once at the end of the day after all work sessions are complete.

Append a work observation entry to today's AI diary file. Use at the end of any session where files were edited or commands were run.

Code review in the voice of a mid-career developer reviewing someone else's PR at 2am, a few beers deep. Use when the user says "酔いどれレビュー", "drunk review", "居酒屋レビュー", "飲みながらレビュー", "酔っ払いレビュー", or asks for a casual code review with personality. Reviews diffs, files, or PR URLs and delivers fragmented, experience-driven reactions as inline comments inside difit via `--comment`.

Test your understanding of what you just learned by teaching it back. Claude plays a curious beginner and asks you to explain concepts from the recent conversation, using the protégé effect. Use when the user says "理解度テスト", "理解度チェック", "quiz me", "test my understanding", "ちゃんと理解できてるかな", or "復習したい".

Learning support agent that adds FIXME hint comments instead of fixing code directly, letting the user implement the fix themselves. Use when the user says "学習モードで", "自分で直したい", "ヒントだけ教えて", "答えは教えないで", "teach to fish".

Check the current implementation diff against a docs/specs spec. Use before committing a completed AC, or when asked to run spec-implementation-check.

Draft or refine a docs/specs spec from a user goal or rough plan. Use when creating a new spec, turning a plan into acceptance criteria, or updating an existing spec.

Implement one acceptance criterion from a docs/specs spec as a transcription of the settled design. Use when starting implementation work on a spec, or when asked to implement an AC.

Information Architect (IA) として情報構造を設計・監査し、サイトマップ／taxonomy／メタデータスキーマ／コンテンツモデル／ナビゲーションモデル／ラベリング仕様を UX/UI デザイナーへのハンドオフ成果物として出力する。発動条件はユーザーが明示的に「information-architect skill を使って」「IA skill で」と、この skill 名を指名したとき、または agent／sub-agent から呼び出されたときのみ使用する

Codex CLI (OpenAI) to consult, review code, or investigate issues. Triggers: "codex", "codexに相談", "codexに聞いて", "codexでレビュー", "レビューして" Use cases: (1) code review, (2) design consultation, (3) bug investigation, (4) problem-solving for difficult issues

Gemini CLI の google_web_search ツールを使ってWeb検索を行う。 Triggers: "検索して", "調べて", "ググって", "web search", "search for" Use cases: (1) 最新情報の取得, (2) 技術ドキュメントの調査, (3) エラーメッセージの検索, (4) ライブラリ・ツールの比較調査